Companies need to forecast sales for two main reasons; the first being to validate medium and long term business plans. Both manufacturing and distribution companies need a minimum amount, and minimum value, of orders coming in to remain viable and, if these orders are not going to arrive, it’s best to know beforehand so that options, like introducing new products or entering new markets, can be considered proactively.
The second reason is that sales increases will impact upon a company’s resources. When a manufacturing company forecasts increased sales, it may have to start planning a new plant well in advance because commissioning new machines and training new staff to operate them will take time. A distribution company might need extra warehouse space and operatives, and their suppliers might also need time to ramp-up production. Conversely, if sales are forecast to decrease, and nothing can be done about it in time, it may be necessary to reduce staff and facilities and doing so in a planned manner may minimize the cost to all concerned.
The problem is that forecasting sales accurately has always been viewed as a difficult thing to do, particularly along complex supply chains, and though demand forecasting software is available, most companies have believed it to be either hideously expensive or, in an attempt to keep the cost down, too simple to be of any real use. Until relatively recently, they were right. But nowadays good, cheap and effective solutions are available, some via the cloud, to link seamlessly with ERP systems to extract data, process it and push forecasts back in.
Manufacturing companies have challenges, but also opportunities, that distribution companies do not. Not only might they need to forecast demand at sellable item level (product or SKU) but they may also need to, or benefit from, forecasting at one or both of the raw materials and intermediate (sub-assembly) levels also.
One problem that some companies have with sales forecasting is the sheer number of products, or variants of products, that they sell. Generating useful forecasts for potentially thousands of products is a task that, for many, would range from the impractical to the impossible and the fact that some items only sell in small numbers adds to the problem. Facilitating useful forecasts in these circumstances is something that ERP, and more-specifically planning bills of materials, can help with.
Planning bills use what we might call “The Law of Big Numbers” and one thing that this says is that big numbers are easier to estimate. As an example, imagine that a car maker has a particular model and that it comes with a number of choices. Customers can have:
- a 1.6 or 2.0-liter petrol or 1.9-liter diesel engine
- a 2 or 4 door, or a station wagon body shell
- in silver, white, red or black paint
- with upholstery in black, grey or brown.
Few car makers are likely to know how many 1.6 liter, four-door cars in red with gray upholstery that they are going to sell next week. But they might have a reasonable idea of how many cars in total that they are going to sell and experience might tell them that overall:
- 20% of the cars that they sell use the 1.6-liter engine, 45% the 2.0, and 35% the diesel,
- 20% will be the 2-door, 65% the 4-door, and 15% the station wagon,
- etc, working down the choices list.
If they create a composite (planning) bill of material that reflects those percentages, and then feed in a forecast at the model level of, say 10,000 cars, the MRP system will tell them that they need:
- 10,000 x 20% = 2,000 1.6 litre petrol engines
- 10,000 x 65% = 6,500 2.0 litre petrol engines
- 10,000 x 15% = 1,500 1.9 litre diesel engines
- 10,000 x 20% = 2,000 2-door body shells
- 10,000 x 65% = 6,500 4-door body shells
- 10,000 x 15% = 1,500 station wagon body shells,
They will never build a car with 0.2 of a 1.6 litre petrol engine, 0.65 of a 2.0 litre petrol engine etc, but the point is that they will have provisioned, reasonably accurately, for actual orders when they come in. From that perspective, it doesn’t matter how all of the options go together on any individual car because that is not important from a planning perspective. But they will know that the items that they have procured are likely to satisfy production demand. By moving the forecast to a higher level, they are dealing with bigger numbers and getting better, safer and more reliable results.
This example has used a car with multiple options as being something easily understood but the basic idea has been applied in manufacturing industries as different as furniture and PC manufacturing. Any company that has subassemblies or intermediates that are used in several end products may gain benefit from it. The key to successful demand forecasting for manufacturing can often be in realizing that it is not always necessary to forecast at the finished item level.
But some manufacturing companies can go further than this because forecasting sales at an intermediate, or subassembly, level opens up the possibility of stocking at that level also and that potentially offers two enormous advantages. Firstly, for make-to-order companies, holding stocks of subassemblies rather than raw materials can significantly reduce manufacturing lead times and that allows some companies to reduce their delivery lead times to customers substantially as well (the difference between production lead time and delivery lead time will be discussed in a future article).
For those that make-to-stock it can remove the frustration of looking in the finished goods warehouse and seeing urgently needed subassemblies that have been built into items for which there are no orders. It’s not always physically possible to disassemble unwanted stock to rob components for high priority orders, even if the required labor can be made available.
Some ERP providers will say that their systems don’t offer planning bills of material, or that holding stocks of things like engines to 2 decimal places is simply too awkward, but a very simple and effective workaround is available when systems don’t have this ability. Using the car example again: a company can set up a part number to describe the model range and give that part number a unit of measure of, say, thousands. Then the bill of material says that to make one thousand cars, two hundred 1.6 litre petrol engines, 650 2.0 litre petrol engines, 150 1.9 litre diesel engines etc are needed and, as long as the forecast is entered in thousands, the system will work just as well.
When using these techniques to generate better sales forecasts, the full power of the MRP element of a company’s ERP system can be let loose and the effect will be well worth the effort.